Considering anisotropy in image reconstruction algorithm for ultrasound computed tomography of trees resulted in a more accurate detection of defects compared to common approaches used.
Context Ultrasound computed tomography is a suitable tool for nondestructive evaluation of standing trees. Until now, to simplify the image reconstruction process, the transverse cross-section of trees has been considered as quasi-isotropic and therefore limiting the defect identification capability.
Aims An approach to solve the inverse problem for tree imaging is presented, using an ultrasound-based method (travel-time computed tomography) suited to the anisotropy of wood material and validated experimentally.
Methods The proposed iterative method focused on finding a polynomial approximation of the slowness in each pixel of the image depending on the angle of propagation, modifying the curved trajectories by means of a raytracing method. This method allowed a mapping of specific elastic constants using nonlinear regression. Experimental validation was performed using sections of green wood from a pine tree (Pinus pinea L.), with configurations that include a healthy case, a centered, and an off-centered defect.
Results Images obtained using the proposed method led to a more accurate location of the defects compared to the filtered backprojection algorithm (isotropic hypothesis), considered as reference.
Conclusion The performed experiments demonstrated that considering the wood anisotropy in the imaging process led to a better defect detection compared to the use of a common imaging technique.
Wood, Orthotropy, Ultrasonic, Wave propagation
Espinosa, L., Brancheriau, L., Cortes, Y. et al. Ultrasound computed tomography on standing trees: accounting for wood anisotropy permits a more accurate detection of defects. Annals of Forest Science 77, 68 (2020). https://doi.org/10.1007/s13595-020-00971-z
For the read-only version of the full text:
The proposed inversion procedure was developed between the years 2015–2018. The numerical codes and datasets are available in the CIRAD dataverse (users may request access to files): https://doi.org/10.18167/DVN1/GI8LSW